The rise of personal AI agents is reshaping the semiconductor market, creating a new opportunity for Advanced Micro Devices and Intel after years of watching Nvidia dominate the artificial-intelligence boom with its powerful graphics processors.
For much of the generative AI era, graphics processing units, or GPUs, have been the star performers. Nvidia’s chips became the industry standard for training and running large AI models following the launch of ChatGPT, pushing CPUs — the traditional workhorses of data-center computing — into the background.
That dynamic is beginning to change.
The emergence of AI agents capable of planning tasks, operating software and working autonomously for hours or even days is increasing demand for conventional central processing units, or CPUs. The shift is already benefiting AMD and Intel, whose processors are designed to handle the sustained computing workloads required by these systems.
“As more agents are created and developed, and more people start to use them for more tasks, it’s going to start to shift the workload away from GPUs and onto CPUs,” said Ryan Shrout, president of Signal65, a consultancy focused on AI hardware and costs.
The change comes as major technology companies race to develop AI agents for consumers and businesses.
OpenAI last week released its AI agent Dots, following Meta’s launch of Muse in early September. Muse gained users rapidly, reaching the top of Apple’s App Store in less than two weeks.
The excitement around AI agents has also translated into the stock market. AMD and Intel had already posted substantial gains this year as demand for CPUs increased, but their shares accelerated further over the past month, rising 32% and 21%, respectively.
AMD’s growing presence in both CPUs and GPUs has helped propel the company into the trillion-dollar market-capitalization club.
AI Agents Need More Than GPUs
The growing importance of CPUs stems from the way AI agents operate.
Unlike traditional chatbots, which generally respond to individual prompts, agents can plan and execute a sequence of tasks with limited human intervention. They may need to access files, run applications, browse the internet, write and test code or perform other computing operations continuously in the background.
That requires a computer environment capable of running those tasks for extended periods.
“CPUs are actually performing the workflows while GPUs are doing the thinking,” said Daniel Newman, CEO of Futurum Group.
GPUs remain crucial for inference — the process of running AI models — but CPUs handle much of the underlying computing work performed by the agent.
OpenAI and Meta rely on virtual machines for their agents. These systems divide physical servers into multiple virtual computers, allowing a single server to support numerous users simultaneously.
Users have also been able to learn something about the hardware behind the agents simply by asking them.
Muse tells users that it operates on an AMD-powered computer, while Dots has indicated that it runs on a virtual computer powered by AMD’s EPYC processors.
Meta, however, says it is not tied to one particular CPU supplier.
“We take a diverse approach to our hardware and are largely CPU-agnostic by design, which gives us the most flexibility in acquiring capacity,” a Meta spokesperson said.
OpenAI likewise uses multiple CPU providers, meaning AMD is competing in a market that includes Intel and other chipmakers.
AMD Emerges as a Major Beneficiary
AMD appears particularly well positioned because of its strong presence among the major cloud providers that operate the infrastructure supporting AI services.
Amazon, Google, Meta and Microsoft all operate large data centers using processors from AMD and Intel, although the cloud giants are also increasingly developing their own chips, many based on technology from Arm.
Arm announced its own CPU designed for AI agents in March, with Meta as its first customer. Arm’s shares have more than doubled since then, reflecting investor enthusiasm for the potential growth of agentic computing.
Nvidia is also moving into the CPU market.
Earlier this year, the company introduced Vera, a newly designed central processor, along with an entire rack built around the CPUs. Nvidia has said the Vera processors were specifically designed for AI agents and projected that the CPU market could reach $200 billion by 2030.
Analysts, however, say Nvidia faces particularly intense competition in CPUs.
“It’s just a new product and they’re going to be competing against a lot of different players,” Jordan Klein, an analyst at Mizuho Securities, said in an interview, adding that he believes AMD currently has the “best product.”
Data Center Growth Accelerates
CPUs are already contributing significantly to AMD’s growth.
The company’s data-center revenue more than doubled to $6.7 billion in the quarter ended in June, representing almost 60% of AMD’s total sales.
AMD executives say the increasing adoption of AI agents is changing the conversation around processors.
″The conversation has changed quite a bit over the last eight to 10 months in terms of agentic usage,” said Dan McNamara, AMD’s senior vice president and general manager of compute and enterprise AI.
McNamara said CPU sales are going to “really ramp.”
One reason is cost.
AMD EPYC processors can contain as many as 192 CPU cores. According to benchmarks and information provided by the AI agents themselves, Muse’s virtual computer uses two cores, while Dots uses nine.
That represents a far less expensive computing environment than the large GPU clusters typically required for advanced AI workloads.
Benchmarks and queries indicate that Dots uses an AMD EPYC 9V74 processor, which can be purchased from resellers for less than $3,000. The EPYC 9D25 reportedly used by Meta for Muse costs even less on the secondary market.
By comparison, a single Nvidia GPU can cost more than 10 times as much, while AI systems typically deploy GPUs in clusters containing hundreds or thousands of chips.
Investors Bet on a Bigger CPU Market
The financial opportunity is attracting increasing attention from investors and analysts.
Meta’s Muse has reportedly surpassed 5 million downloads since its launch last month, according to Sensor Tower. That growth has raised questions about how much it will cost Meta to operate the service as usage increases.
Morgan Stanley estimates that serving Muse could cost between $3 and $130 per month per user, with an average cost of approximately $37, depending on how heavily users rely on AI inference.
The investment bank estimates that Meta’s AI agent could account for 20% of AMD’s chip sales in 2026.
Industry forecasts for the broader CPU market are also becoming more optimistic.
Futurum estimates that global CPU sales could reach $118 billion in 2027, nearly twice the firm’s previous forecast issued in May.
AMD has been even more bullish. In July, the company predicted that the overall CPU market could reach $220 billion in annual sales by 2030, sharply raising its previous 2025 forecast of $60 billion.
AMD also said it expects to capture more than half of that market.
Mercury Research reported in August that AMD controlled about 46% of the x86 CPU market by unit shipments, putting the company in a strong position as demand for processors increases.
Intel, meanwhile, is also benefiting from the renewed demand.
A New Chip Battle
The resurgence of CPUs does not mean GPUs are becoming obsolete. AI models still require powerful GPUs for training and inference, and Nvidia remains the dominant force in that market.
Instead, the growth of AI agents is expanding the amount of computing required around those models.
For AMD, that could be particularly valuable because the company can supply both CPUs and GPUs. Its x86 architecture also gives it an advantage in data centers where customers want compatibility with existing software and infrastructure.
“AMD is probably getting the most of this because they have the largest market share and the cloud hyperscaler world for CPUs,” Klein said, adding that he “wouldn’t rule Nvidia out.”
The emerging market is therefore becoming less about choosing between CPUs and GPUs and more about how the two types of processors work together.
“If you are a company that can supply both types of chips, you’re in the best position,” Klein said.
As AI agents move from experimental tools to everyday digital assistants capable of performing increasingly complex tasks on their own, the quiet work performed by CPUs could become just as important to the AI economy as the high-powered GPUs that launched the current revolution.